JTF (just the facts): Five photo-based artworks—one framed pigment print; one framed dye sublimation metal print, one set of ten individually framed pigment prints; one set of four individually framed pigment prints; and one print, comprising hundreds of images, on adhesive wall material—hang on the white walls of the gallery’s front room. A 12-minute, single-channel color video plays in a viewing room between the gallery’s front and back rooms. Eleven framed dye sublimation metal prints hang on the white walls of the gallery’s back room. An additional framed dye sublimation metal print hangs on a white wall in the gallery’s foyer. The photographic works range in size from 21 1/2 x 26 7/8 inches to 193 x 55 1/8 inches (the wall sticker). All of the works were made in 2017 and are available in an edition of five, with the exception of the wall sticker, which is unique, and the video, which is available in an edition of three. (Installation shots below.)
Comments/Context: In 1998, a year after graduating from UC Berkeley, Trevor Paglen embarked on his first study of invisible places; posing as a student, he visited California penitentiaries wearing a hidden microphone for a project he called “Recording Carceral Landscapes.” In 2000, after earning an MFA from the School of the Art Institute of Chicago, he went back to the University of California to study geography. In the course of his researches, he noticed blank spots on aerial photographs of the Western United States and realized that these redacted portions represented the locations of top-secret military installations. In 2003, he started taking pictures of them.
Over the next dozen or so years, Paglen traveled to and shot blurry photographs of places most people are never even aware exist: test sites, CIA secret prisons, and “choke” points where communications cables come close enough to shore to be tapped (and, it is rumored, are). His last show at Metro Pictures in 2015 included grainy underwater images of cables running over the ocean floor; video footage, taken from a distance, of an NSA surveillance station in the UK; and even a computer that visitors could use to access the Tor Network, which enables anonymous communications.
For his current show at the gallery, Paglen has turned his attention to invisible images—specifically, the pictures created or processed by machines, from self-driving cars to Facebook’s artificial intelligence networks, which generally only exist as code. Working with computer-vision and artificial intelligence researchers at Stanford University, he has created a three-part installation exploring what he calls “machine vision” and the biases built into its supposedly neutral algorithms.
The first part of the exhibition centers on the process of teaching AI software to recognize objects, faces, and activities, for which the AI is given libraries of images called training sets. One work is a “faceprint”—a record of the physical characteristics unique to a particular individual—of the philosopher Frantz Fanon, translated from code into a readable image. To make a faceprint, an AI needs many pictures of the subject—which are usually easily found on social media. Nearby is a wall of pictures of the artist Hito Steyerl, who smiles, pouts, and scowls for the camera; each picture is accompanied by a determination on her emotional state made by a facial-analysis algorithm.
A video playing in a central viewing room presents, in rapid-fire sequence, two kinds of imagery. Video clips of people engaged in various activities are interspersed with digital images representing the AI’s analysis of what it is seeing. The clips show humans laughing, crying, hugging, pushing cars, and doing chin-ups; the digital images show the same subjects broken down into black-and-white abstractions.
The final, and most provocative, portion of the show asks what might happen if an AI were given training sets based on alternative taxonomies. For the images here, collectively titled Hallucinations, Paglen and his cohorts created “irrational” sets based on literature, folklore, and poetry, and asked a pair of AIs to use them to create pictures of such things as a vampire, a post-human landscape, and a predator (the training set for this image included pictures of carnivorous plants, drones, and Mark Zuckerberg).
The exercise vividly demonstrates that while machines are neutral, their programmers aren’t. As Paglen told Artforum in July, “These are not merely representational systems or optimization systems; they are set up as normative systems and therefore they become enforcement systems.” At the same time, however, the artistic efforts of Paglen’s AI are painterly, surreal, and quite beautiful. While the show ostensibly investigates what it is to be a machine, its real subject is what it means to be human.
Collector’s POV: The works in this show are priced between $20000 (for the smallest framed print) and $60000 (for the video). Paglen’s work has not yet reached the secondary markets with any regularity, so gallery retail likely remains the best/only option for those collectors interested in following up.
The ‘AI’ generated ‘hallucination’ images are pretty extraordinary. They may be software-output but appear convincingly organic – and unsettling as such. The fact people have initially constructed algorithams and then edited the output for art-piece interest does not quite diminish the sense that we are witnessing something totally unfamiliar and authentic, and highly distinct from other avenues of digital manipulation being investigated.